Permalink
Commits on Jun 11, 2011
  1. Checking in changes prior to tagging of version 0.06.

    Changelog diff is:
    committed Jun 11, 2011
  2. Update Manifest

    committed Jun 11, 2011
  3. Prepare for a 0.06 release

    committed Jun 11, 2011
  4. Replace 'perl' by $^X in test

    RT #68215. Thanks Andreas Koenig
    committed Jun 11, 2011
  5. Merge branch 'fixes'

    committed Jun 11, 2011
  6. Guaranty monotonically increasing seq at sec granularity

    Fix a problem where the yuidd would just increment previous
    id (just incrementing the serial) until exhausting the serial before
    refreshing the timestamp. This prevent two ntp synced server
    to generate ordered id at the second granularity
    committed Jun 11, 2011
Commits on Jun 10, 2011
  1. Add debug utils and more tests

    committed Jun 10, 2011
  2. Remove trailing spaces

    committed Jun 10, 2011
Commits on Jan 12, 2011
  1. Adds a connect timeout configuration option.

    The new default is to wait for 1s. Before we would block enter
    a tight loop if all servers are down.
    
    Also adds documentation for the two timeouts.
    committed Jan 6, 2011
Commits on Mar 23, 2010
  1. Makes the build a bit better

    committed Mar 23, 2010
  2. Makes Data::YUID work on a 32bit system

    It's slower.
    committed Mar 23, 2010
Commits on Jun 11, 2007
  1. Notes for 0.03

    Kevin Goess committed Jun 11, 2007
  2. - got rid of a warning at reconnect

     - create a new connection every 1,000+rand(1,000) connections, to force rebalancing
       if a server goes down and comes back
    Kevin Goess committed Jun 11, 2007
  3. Checking that Active_Sock is actually open.

    Kevin Goess committed Jun 11, 2007
Commits on Jun 5, 2007
  1. - The original behavior is that the client would create a connection to

           every server it knows about on new().  Those connections would be chached,
           but the next new() call would create four new connections anyway.  So each
           YUID server was getting a persistent (but not re-used) connection from
           every (TypePad) app server, and they were choking on the sheer number of
           connections.
     
           (There was a suboptimal thing happening in connect_to_servers(), which
           ignored the connections in the cache, if any.)
     
           So now, we connect only when a get_id request comes in, and only
           to one server. If that server can't get us an ID, we try the
           other ones. Each server will only have to maintain N/S connections,
           instead of N (N=number of app servers, S=number of YUID servers).
    Kevin Goess committed Jun 5, 2007
Commits on Dec 1, 2005
  1. script to benchmark vs database native numbering

     (is this a realistic thing to measure?)
    Mark Paschal committed Dec 1, 2005
Commits on Nov 23, 2005
  1. Added a Nagios plugin to monitor a yuidd server

    Ben Trott committed Nov 23, 2005
Commits on Nov 22, 2005
  1. Simplified protocol to just an OK for a valid response, with optiona…

    …l args.
    
     Added "shutdown", "ping", and "stats" commands (very simple for now).
    Ben Trott committed Nov 22, 2005
  2. Added more robust isgoodid test, along with a test where we kill off…

    … one of the servers.
    Ben Trott committed Nov 22, 2005
  3. Modified to use non-blocking I/O, and handle failures on a connected

     socket by switching to use a different server.
    Ben Trott committed Nov 22, 2005
  4. Initial checkin

    Ben Trott committed Nov 22, 2005